Face Recognition

Use:

"Face Recognition System" is intended for identification of persons known to the system regardless of any face changes connected with age, hair style and hair colour changes as well as light conditions.

Specifications:

Algorithms based on the analysis of biometric characteristics cannot high result of identification. This result is lower than 50% under uncontrolled conditions and without a special and very long training. Smilart Ltd has developed and has been constantly improving innovative recognition technologies which are based on reconstruction of face surface from ordinary images and identification of individual characteristics.

Smilart technologies provide high accuracy, speed and stability of identification under difficult conditions:

• different positions of the head;
• a large number of persons in the image or video;
• the presence or absence of mustache, beard, glasses, changes in hair style;
• age-related changes;
• a large number of persons kept in the system base;
• uncontrolled conditions: illumination, brightness, contrast, different backgrounds.

At the moment the probability of finding all the images of a particular person in the base of 3000 individuals provided one image per person comparison (a stably illuminated neutral full-face image is desired) without separation to sets and classes is in the interval from 0,8– to 1.0, i.e. the comparison type is 1 to 1.

The probability of recognition is close to 1.0 if there is a large number of images and they are rather different (head position, mimics, age, illumination). In this case the system is trained through object images.

System testing:

Uploaded images and those in the database were taken in different periods of time (difference is up to 5 years). Controlled and uncontrolled illumination was used. The images were taken not in a studio, but in natural conditions by amateur photographer using nonprofessional or semiprofessional equipment (in this case the result is close to 1.0).

At the moment Smilart is finishing the test of modules responsible for iris and fingerprint recognition. Soon "Identification and Verification System" is going to support these methods of recognition.

Supported camera manufacturers:

• Arecont Vision;
• GANZ;
• Panasonic;
• Axis;
• Wi-Fi cameras of different manufacturers;
• Beward;
• Web cameras of different manufacturers;
• Basler;
• Sony;
• Bosch;
• Delco.

It's possible to receive images from IP-cameras of other manufacturers if they support JPEG or MJPEG format transfer by HTTP protocol or H.264 format transfer by RTPS protocol. It's necessary to know URL (request line) of required resource.

Note: camera URLs of different manufacturers are different. Ask your provider or manufacturer for URL data or specify the possibility to acquire this information in advance.